Alternative bridge Spoke Permanent Magnet Synchronous Generator Design for Wind Power Generation Systems

Generators are a key technological element of wind power generation systems. The use of synchronous generators that employ heavy rare-earth permanent magnets is increasing owing to the demand for higher efficiencies. However, these magnets are expensive and undergo an unbalanced supply and demand. Therefore, spoke-type structures have been developed and used in various fields to compensate for the deficient performance when using a ferrite magnet, which is inexpensive and provides a stable supply and demand. However, permanent magnet synchronous generators (PMSGs) with adopted spoke-type structures and an output power of 1 kW or higher have not been studied and practically applied sufficiently. Additionally, new designs must consider differences between the target specifications of a generator and wind turbine. In this study, we designed and analyzed the characteristics of an alternative bridge spoke-type PMSG modeling a generator for a 3-kW class wind turbine. We analyzed the performance of the existing spoke-type generator and proposed an alternative bridge spoke shape with an improved performance by removing the leakage flux of the bridge. The performance of the existing spoke model was compared using the finite element method (FEM), and the performance of the proposed model was verified more accurately by performing a three-dimensional (3D) analysis considering the 3D effect. Additionally, we analyzed the voltage fluctuation rate according to the number of stator slots and overhang structure and designed a final model. Considering the characteristics of the ferrite magnet used in the spoke-type design, the irreversible demagnetization characteristics of the conventional and final models were compared using simulations, followed by a comparison between the stiffness characteristics of the two models. Finally, a prototype was manufactured, and the feasibility of the final design was verified using the performance tests.


I. INTRODUCTION
Recently, interest in and studies on renewable energy, particularly on ecofriendly and economic wind power systems, have been continually growing [1][2][3][4][5][6]. Generators are a key technological element of a wind power generation system because their performance can directly reduce the weight and increase the efficiency of the corresponding wind power generation system. Many generators, such as switched reluctance and induction generators, have been used; however, the use of permanent magnet synchronous generators (PMSGs) is increasing owing to their high output density per unit volume and the possibility to be miniaturized [7][8][9][10]. The permanent magnet types employed in PMSGs are ferrite and heavy rareearth magnets. Heavy rare-earth magnets with a high energy density are preferred because of their high efficiency; however, their supply and demand are unstable and expensive [11][12]. The supply and demand of ferrite magnets are stable in addition to their lower cost compared with that of rare-earth magnets. However, ferrite magnets have a low magnetic flux density, which is approximately one-third that of rare-earth magnets, making it difficult for them to achieve the same performance as rare-earth magnets. Therefore, a spoke-type structure was developed to overcome the problem of low magnetic flux density. Therefore, spoke-type structures have been widely studied and commercialized in permanent magnet synchronous motors (PMSMs) [13][14][15][16][17]. In particular, many studies are being actively conducted on the spoke-type PMSMs for different applications, such as shape studies to improve demagnetization or magnetization performance, and performance maximization using an asymmetric structure and barrier design in the rotor core, use of magnetic materials applied with new materials [18][19][20][21][22][23][24][25]. However, the research and development and application cases of the spoke-type structure are still insufficient for 1-kW or higher generators [26][27][28][29][30]. Although structures of the spoke-type INSIDabcdef_:MS_0001MS_0001 PMSMs and a PMSGs are similar, the required specifications are distinct. In particular, the cogging torque, voltage fluctuation rate during load operation, and efficiency for wind turbines are extremely important design specifications. In addition, the cogging torque is directly associated with the starting performance. Therefore, the voltage fluctuation rate and cogging torque must be reduced for a spoke-type PMSG with a large leakage magnetic flux in the axial direction. In this study, we investigated the design and characteristics of an alternative bridge spoke-type PMSG for use in wind turbines. The alternative bridge spoke-type PMSG is designed using a 3-kWclass PMSG as a model. Subsequently, the performance of an existing spoke-type and that of the proposed alternative bridge PMSG were compared using the finite element analysis (FEA) method. Additionally, an accurate 3D analysis was conducted considering three-dimensional (3D) effects, such as the spoketype axial leakage magnetic flux, and the performance changes were examined. The proposed model was designed by analyzing the stiffness according to the number of stator slots and overhangs for the voltage drop characteristics. Table 1 demonstrates that the characteristics of the proposed alternative bridge spoke-type PMSG are superior to those of the existing permanent magnet rotor shapes. Considering the no-load and load characteristics of the proposed model and those of the permanent magnet material (ferrite), the irreversible demagnetization and stiffness characteristics of the existing and proposed models were compared using simulation. Finally, prototypes were manufactured, and performance tests were performed to verify the validity of the proposed design and characterization.

A. PMSG principle and equivalent circuit
The PMSG has the same structure as the PMSM; however, a difference exists in whether the input is mechanical or electrical energy. As the input mechanical energy rotates the generator shaft, the magnetic flux generated by the permanent magnet in the rotor generates an induced electromotive force in the armature coil. In addition, when stator is connected to a load, current flows. In PMSGs, the no-load electromotive force E0 arising from the magnetic flux generated by the permanent magnet and the armature reaction electromotive force ea generated by the armature reaction magnetic flux are induced. Furthermore, ea can be obtained as follows [31][32][33]: where λa is the armature reaction flux linkage, ϕa is the armature reaction flux, and N is the number of turns. Equation 1 can be expressed as the armature reaction electromotive force phasor Ea in the frequency domain as follows: a a a E jX I = (2) where Xa is the armature reaction reactance. If leakage flux exists in a PMSG, the induced electromotive force El is expressed as follows: where where Xl is the leakage reactance. In Equation 4, the sum of Xa and Xl is called the synchronous reactance Xs, and the sum of Xs and Ra is called the synchronous impedance Zs.
The voltage V on the terminal side of the generator is obtained by subtracting the voltage drop owing to the synchronous reactance Xs and the voltage drop owing to the winding resistance Ra from the no-load induced electromotive force E0, which is the same as that expressed in Equation 5.  In the rotor that generates the no-load counter electromotive force E0 of the equivalent circuit, the magnetic arrangement torque acts in the opposite direction to the rotational direction of the field. Therefore, a power is needed to overcome the selfarranged torque generated in the direction opposite to the rotational direction and return it to a constant speed. Equation Moreover, Ploss is the sum of the electrical and mechanical losses, and the torque applied to the rotor is expressed as  Table 2 lists the target specifications of a PMSG for a 3-kWclass wind turbine. Fig. 2 shows a 10-pole 12-slot PMSG using a conventional spoke-type rotor designed using the voltage and output power equations.  The conventional spoke-type rotor consists of a permanent magnet and a pole piece, an inner bridge to connect it to the shaft, a position-locking projection to fix the position of the magnet, and an outer bridge to prevent the magnet from scattering. Fig. 3(a) shows a conventional spoke-type magnetic circuit, where Φg is the airgap magnetic flux, Φm is the magnetic flux, and Φs and Φr are the magnetic fluxes of the stator and rotor, respectively. Rm is the magnetic resistance of the magnet, Rs and Rr are the magnetic resistances of the stator and rotor, respectively, and Rg is the magnetic resistance of the air gap. Φlo, Φli, and Φlp are the leakage fluxes flowing through the outer, inner, and position locking projections, respectively. In addition, Rlo, Rli, and Rlp are the corresponding magnetic resistances. Each bridge is structurally necessary because it fixes the pole piece and the magnet, and prevents scattering; however, it reduces the electromagnetic performance because of the generation of a leakage magnetic flux, as shown in Fig.  3(a). In addition, the magnetization performance of a spoketype rotor is important, and the position locking projection interferes with the magnetization of the lower part of the magnet. In general, the inner and outer bridges are designed to be as narrow as possible to induce magnetic saturation, which helps reduce the leakage magnetic flux. Concurrently, as the thickness of a bridge decreases, the rigidity becomes disadvantageous during operation; thus, the bridge thickness reduction is limited.

. Magnetic equivalent circuits (a) conventional and (b) alternative bridge spoke models
Rm is the magnetic resistance of the magnet, Rs and Rr are the magnetic resistances of the stator and rotor, respectively, and Rg is the magnetic resistance of the air gap. Φlo, Φli, and Φlp are the leakage fluxes flowing through the outer, inner, and position locking projections, respectively. In addition, Rlo, Rli, and Rlp are the corresponding magnetic resistances. Each bridge is structurally necessary because it fixes the pole piece and the magnet, and prevents scattering; however, it reduces the electromagnetic performance because of the generation of a leakage magnetic flux, as shown in Fig. 3(a). In addition, the magnetization performance of a spoke-type rotor is important, and the position locking projection interferes with the magnetization of the lower part of the magnet. In general, the inner and outer bridges are designed to be as narrow as possible to induce magnetic saturation, which helps reduce the leakage magnetic flux. Concurrently, as the thickness of a bridge decreases, the rigidity becomes disadvantageous during operation; thus, the bridge thickness reduction is limited. Unlike the conventional spoke-type structure, in the alternative bridge spoke-type structure, only half of the inner and outer bridges are used, and the scattering of the pole pieces or the permanent magnet is prevented in the intersection of the inner and outer bridges. In addition, by making an angular inside for the rotor and developing an optimal design, the magnet was fixed without a position-locking projection. The corresponding alternative bridge spoke-type magnetic circuit is shown in Fig. 3(b), where Rgo and Rgi are the magnetic resistances of the outer and inner airgaps, respectively, and Φgo and Φgi are the corresponding leakage fluxes.

C. Alternative bridge spoke-type design
The two-dimensional (2D) FEA was conducted to validate the performance of the proposed alternative bridge spoke model, and its performance was compared with that of the conventional spoke-type model. To this end, the outer and inner diameters, and air gap of the stators and rotors, and the number of magnets used in both models were identical.  The alternative bridge spoke model significantly reduced the leakage magnetic flux owing to the conversion of half of the outer bridges into voids, as shown in Fig. 3(b). Similarly, half the number of inner bridges exist, as shown in Fig. 3(b). Moreover, fixing the magnet without a supporting structure blocked the leakage magnetic flux generated in the positionlocking projection. A decrease in the leakage magnetic flux increased the back electromotive force between the no-load lines of the alternative bridge model by approximately 4.0% compared to that in the existing spoke-type model. Furthermore, total harmonic distortion (THD) of the line back EMF improved from 9.6 to 4.48%. Because the number of outer bridges was reduced by half and the thickness was minimized in the alternative bridge model, changes in the reluctance were also minimized, reducing the cogging torque by approximately 47.1% compared to that of the existing spoke-type model. Table 3 presents a performance comparison of the two models for power generation. The alternative bridge spoke model increased the power generation voltage by approximately 5.3% when the output is 3 kW, which could be attributed to the reduction in both the leakage flux and current owing to the removal of the bridges. In addition, because current reduction reduces copper loss, the alternative bridge spoke-type model exhibited an increase of 0.2% efficiency compared to the conventional model. Moreover, the voltage fluctuation rate of the alternative bridge mode, which is an significant specification, was also improved over the existing model.

A. Three-dimensional modeling and performance analysis
An accurate 3D performance analysis was conducted considering 3D effects, such as the rotor axial and end coil leakage fluxes generated in a spoke structure, and the fringing effect in the airgap. Fig. 7 illustrates a schematic of the 3D shape of the alternative bridge spoke-type model.  The results of the 3D and 2D FEA of the alternative bridge spoke model are summarized in Table 4. The 2D and 3D analysis results were similar under the no-load condition. However, the 2D and 3D analysis results differed during the power generation operation because the 3D effects were not reflected in the 2D analysis. Moreover, a large voltage drop occurs during power generation owing to the 3D effects, and the voltage and voltage fluctuation rates are not satisfied, as defined in the target specifications.
In PMSGs, electric current flows in the stator coil during power generation, and the resulting magnetic flux of the armature reaction flows through the stator core, airgap, and rotor core. The magnetic path between the stator and rotor has a magnetic flux component in the radial direction and an axial magnetic flux component generated by the fringing effect. The axial magnetic flux component, which cannot be considered in the 2D analysis, is reflected in the 3D analysis, resulting in a reduction in the total magnetoresistance of the armature reaction. Therefore, during load operation, the magnitude of the armature reaction in the 3D analysis (and not in the 2D analysis) increases. Therefore, the voltage generated during the load operation decreases in the 3D analysis compared to that of the 2D analysis because of an increase in the magnitude of the armature reaction. The armature reaction is an inductance component in a self-equivalent circuit, and the inductance is proportional to the square of the number of turns. Therefore, an additional design to reduce the number of turns is required to improve the voltage drop owing to the armature reaction.

B. Reducing voltage drop
We developed a design to reduce the number of turns required to meet the target specifications by reducing the voltage drop. First, we increased the number of stator slots, changed the winding method and the number of turns, and examined the trend. We examined 12, 24, 36, and 48 stator slots, for each of which the phase resistances and number of turns are listed in Table 5.     Fig. 9 shows the copper losses and efficiencies. The number of turns decreases as the stator slots increase for an output of 3 kW. The magnitude of the armature reaction decreases as the number of turns decreases. As the number of stator slots increases, the voltage fluctuation rate decreases. Conversely, the magnitude of the voltage increases during load operation. When the number of slots increases, the size of the phase resistance increases owing to an increase in the number of end turns, and the current decreases, thereby increasing the efficiency. Therefore, 36 and 48 slots satisfy the voltage fluctuation rate, as specified in the target specifications. Thus, the 36-slot model with an excellent efficiency and manufacturability was selected. Rotor overhangs of 1.25, 3.5, and 7 mm were applied as one-sided standards, and the trend was determined by conducting 3D FEA. Fig. 10 shows the voltage fluctuation rate and efficiency variation with the overhang length.

A. Comparison of the demagnetization characteristics
The current generated by the wind creates an armature reaction flux and flows toward the load. A current larger than the rated current may flow owing to the size of the load resistance or failure. Thus, the magnitude of the armature reaction force and magnetic flux increases in proportion to the current, which affects the irreversible demagnetization of the magnet. Fig. 12 shows that the operating point of the permanent magnet device is determined at the point where the B-H curve of the permanent magnet and the PC line intersect.  The PC line is determined by the shape of the permanent magnet machine as follows: where Bm = ϕ/Am, Hm = Fm/lm, lm is the magnet length, Am is the magnet area, Ag is the airgap area, and g is the airgap length. In addition, where Fm is the magnetomotive force. As the current increases, the component of the armature reaction increases, and the Pc line moves to the left. Irreversible demagnetization occurs when the operating point is formed below the knee point.
In particular, ferrite magnets have a low coercive force compared to rare-earth permanent magnets (NdfeB, SmCo); thus, they are vulnerable to demagnetization by a reverse magnetic field, where demagnetization occurs at low temperatures [34][35][36][37]. The current density of the proposed model is approximately 2 A/mm 2 and the operating temperature is in the range of 20-50 °C; thus, permanent magnet irreversible demagnetization by temperature does not occur. Subsequently, the demagnetization characteristics of the existing model and the proposed model owing to the armature reaction were compared using simulation. The demagnetization current owing to a fault is twice that of the existing rated current. Fig. 13 illustrates the irreversible demagnetization area according to the shape of the rotor. The amount of permanent magnet used in both models is approximately the same; however, it can be confirmed that irreversible demagnetization of the conventional spoke model occurred in some areas of the permanent magnet. Unlike the conventional spoke model, an irreversible demagnetization area did not occur in the proposed alternative bridge spoke-type model by removing the external bridge or applying a barrier. This is because the removal of outer bridge and the application of an air gap barrier makes the magnetic resistance on the outer diameter side of the rotor be higher than that of the conventional spoke model, which leads to be more robust to reverse magnetic fields. The mechanical stiffness characteristics during road load operation are compared because the shapes of the existing and proposed models are different. The mechanical stiffness characteristics are compared with the maximum value of the equivalent stress that can occur in the model and the safety factor. The maximum value of the equivalent stress indicates the maximum stress that can occur in the model. The safety factor is an index that guarantees structural stability against uncertainties caused by various factors. Although the safety factor standard varies depending on the design, it usually has a value of 2 or more, as expressed in Equation (12).

B. Comparison of the mechanical rigidity
Fig. 14 presents a comparison between the rigidity of the conventional and final models. The equivalent stresses of the conventional and proposed models are 249.43 and 368.52 kPa, respectively. Therefore, the final model is disadvantageous in terms of the mechanical rigidity. However, the safety factor is 15 for both models. As the safety factor of the final model is similar to that of the existing mass-produced model, the rigidity of the final model has no problem, and thus no abnormality exists in the production result.

V. EXPERIMENTAL WORK AND RESULTS
A prototype generator was manufactured to verify the design and performance of the designed generator, and its performance was examined. Fig. 15(a) shows the manufactured stator and rotor, and Fig. 15(b) presents the test configuration of the generator prototype and dynamometer.
(a) Generator prototype stator and rotor core cross-section (b) Test systems including generator prototypes and dynamo meters

FIGURE 15. Photographs of Generator prototype core and test system
In the tests, the no-load back electromagnetic force, cogging torque, voltage, current, and torque at a rated operating speed of 300 rpm were measured. Fig. 16 presents the measured waveform using an oscilloscope and the no-load back EMF waveform analyzed using FEA. According to the comparison results, the error in the size of the effective value was approximately 5%, indicating a high accuracy.  Table 7 presents a comparison between the experimental and simulation results for the prototype. The test was performed at an environmental temperature of 20°C, and the Fig.17 is a graph measuring the temperature rise during load operation. The temperature sensor is attached to the stator winding side, and due to the low current density, it can be seen that the temperature saturation appears at about 30℃. The load resistance was 14.3 Ω, and the resulting measured output power was 3.16 kW. Table 7 confirms that most of the performance results during the no-load and load operations are comparable with the FEA results. However, the current was slightly higher than that of the FEM analysis, which slightly increased the output based on the load resistance setting during the test. Therefore, current and torque were slightly increased. In addition, the efficiency decreased owing to an increase in the iron loss, which is attributed to the harmonic iron loss and analysis error.
Except for the above cases, the simulation and experimental results were comparable, validating the design and performance of the proposed alternative bridge spoke model. This study performed a design and characteristic analysis of a alternative bridge spoke-type PMSG. In contrast to the conventional spoke type, a crossroad bridge was applied, and the support structure was eliminated to minimize the leakage magnetic flux. Therefore, a higher line voltage and lower cogging torque than those of the existing spoke-type model were achieved. In addition, 3D FEA analysis confirmed that the voltage fluctuation rate was higher than that of the 2D analysis owing to the 3D effect, and the final model, to which an overhang was applied, was selected by analyzing the performance trend based on the number of stator slots and overhang. This is to reduce the voltage drop owing to the armature reaction because the armature reaction component is proportional to the square of the number of turns. The final model was selected by changing the number of turns according to the number of slots and reducing the number of turns using overhangs to improve the voltage drop and examine other performance parameters, such as current and efficiency. The final model improved the voltage fluctuation rate by more than 8% compared to the existing model and satisfied all target specifications. In addition, considering the characteristics of permanent magnets, irreversible demagnetization was compared using simulation. It was confirmed that the proposed model is stronger in irreversible demagnetization than the existing model. Considering the changes in the rotor shape, the mechanical stiffness was also compared with that of the existing model. Although the maximum stress was slightly increased, the stiffness characteristics were comparable to those of the existing model, and the safety factor was checked to confirm that there was no problem in mechanical stiffness even during load operation. Finally, a prototype was manufactured, and a performance test was conducted to verify the validity of the design and analysis results. The results obtained in this study will be helpful for accurate characterization and design of low-and medium-capacity generators. In addition, the proposed model is expected to contribute toward applied research in various fields.